Direct Model Generation for Subject-specific on-segmented Medical Volume Data

Segmentation is an important prerequisite among many conventional modeling techniques in producing a surface or volume mesh. This poses as a major obstacle in the surgical simulation work flow due to the potential processing time and human intervention involved. The procedure's impracticability hinders subject-specific surgical applications from deploying in daily clinical environment. In this work, a direct, segmentation-free model generation method is proposed. The two-step method generates a volumetric model from a set of unprocessed, non-segmented medical volume data. With the use of “Volume Proxy Mesh” (VPM) as the foundation mesh structure, the entire volume data is converted into a tetrahedral mesh in the first phase. Specifically, vertices in the mesh are directly placed to sample the geometric details in the dataset based on each voxel’s curvature information. Thus, segmentation is strategically lessened to trivial automated step. In the second phase, Particle Swarm Optimization (PSO), a global optimization algorithm, is employed to maximize the deformation quality of the initial resulting mesh. Through a designed cost function, the initial mesh quality is improved iteratively until criteria are reached. Visualization results of several sets of subject-specific kidney data illustrate the effectiveness of the proposed algorithm and the applicability of PSO in tetrahedral mesh refinement.